A new study indicates large language models can link anonymous social media accounts to real-world identities by correlating seemingly innocuous clues across platforms, sharply lowering the cost and expertise needed for privacy attacks. Researchers warn the technique could fuel state surveillance and highly targeted scams, even as errors and false positives remain a risk. Academics caution that public datasets beyond social media—such as hospital or admissions records—may no longer meet anonymization standards in the AI era. Suggested mitigations include tighter platform controls on data access and user restraint in sharing identifiable details.
Related articles:
Fitness tracking app Strava gives away location of secret US army bases
De-anonymizing Social Networks
